আপ করুন এবং নিচে, আপ করুন এবং নিচে

Question

চ্যালেঞ্জ:

ধনাত্মক পূর্ণসংখ্যা ইনপুট এন দেওয়া হয়েছে , এই ভ্যাটারি অনুসরণ করে এমন একটি ভেক্টর তৈরি করুন:

0  1  0 -1 -2 -1  0  1  2  3  2  1  0 -1 -2 -3 -4 -3 -2 -1 ... ±(n-1) ±n

বা, শব্দের মাধ্যমে ব্যাখ্যা করা হয়েছে: ভেক্টরটি শুরু হয় 0এবং 1ক্ষুদ্রতম বিজোড় ধনাত্মক পূর্ণসংখ্যার না পৌঁছানো পর্যন্ত বৃদ্ধি করে যা ক্রমটির অংশ নয়, তারপরে এটি ক্ষুদ্রতম (প্রস্থে) এমনকি নেতিবাচক পূর্ণসংখ্যার অবধি পৌঁছা অবধি কমিয়ে দেয় অনুক্রমের অংশ নয়। এটি nপৌঁছানো অবধি এভাবেই চলতে থাকে। ক্রমটি বিজোড় nহলে ইতিবাচক nএবং সমান nহলে নেতিবাচক হবে n।

আউটপুট বিন্যাস নমনীয়।

পরীক্ষার কেস:

n = 1
0  1
-----------
n = 2
0  1  0 -1 -2
-----------
n = 3
0  1  0 -1 -2 -1  0  1  2  3
-----------
n = 4
0  1  0 -1 -2 -1  0  1  2  3  2  1  0 -1 -2 -3 -4
-----------
n = 5
0  1  0 -1 -2 -1  0  1  2  3  2  1  0 -1 -2 -3 -4 -3 -2 -1  0  1  2  3  4  5

আপনি শূন্য সূচকে এন নিতে বেছে নিতে পারেন । n = 1তাহলে দিতে হবে 0 1 0 -1 -2।

এটি কোড-গল্ফ , তাই প্রতিটি ভাষার সংক্ষিপ্ততম কোডটি জয়ী হয়! ব্যাখ্যা সর্বদা হিসাবে উত্সাহিত হয়!

Comments

প্রাসঙ্গিক: OEIS A196199 ।

— মিঃ এক্সকডার

Answer 1

আর , 58 54 50 48 43 bytes

-২ বাইট ধন্যবাদ মিকিটি-কে ধন্যবাদ

function(n)diffinv(rep(1:n%%2*2-1,1:n*2-1))

এটি অনলাইন চেষ্টা করুন!

function(n)
 diffinv(                           # take cumulative sum, starting at 0 of
             1:n%%2*2-1,            # a vector of alternating 1,-1
         rep(                       # repeated
                        1:n*2-1))   # 1, 3, 5, etc. times

Answer 2

Perl 6,  60  26 bytes

{flat {((1,-*...*)ZX*(-$++...0...$++)xx$_)}(),$_*($_%2||-1)}

Try it

{[...] (-1,-*...*)Z*0..$_}

Try it

Expanded:

{  # bare block lambda with implicit parameter $_

  [...]  # reduce using &infix:«...» (sequence generator)

          ( -1, -* ... * ) # (-1, 1, -1, 1 ... *)

      Z*                   # zip multiplied with

          0 .. $_          # range up to and including input
}

(-1,-*...*)Z*0..$_ generates the sequence 0 1 -2 3 -4 5

Answer 3

Python 2, 69 57 56 bytes

f=lambda n:[0][n:]or f(n-1)+range(-n,n+1)[::n%2*2-1][2:]

Try it online!

For each n up to the input the range(-n,n) (inclusive) is calculated, inverted when n is an even number, has the fist two numbers (after the inversion) removed, and then appended to the output.

Answer 4

05 এ বি 1 ই , 9 7 বাইট

Saved 2 bytes thanks to @Emigna

Ýā®sm*Ÿ

Try it online!

My first 05AB1E answer (I think), so I may be missing some tricks...

Explanation

Ý         # push range [0 ... n]   stack: [[0 ... n]]
 ā        # push range [1 ... len(prev)]  [[0 ... n], [1 ... n+1]]
  ®       # push value of register        [[0 ... n], [1 ... n+1], -1]
   s      # swap top two values           [[0 ... n], -1, [1 ... n+1]]
    m     # power                         [[0 ... n], [-1, 1, -1, 1, ...]]
     *    # multiply                      [[0, 1, -2, 3, -4, 5, ...]]
      Ÿ   # range interpolation           [[0, 1, 0, -1, -2, -1, ...]]

I have to thank @Dennis for the original usage of Ÿ, otherwise I may not probably would never have known about it...

Comments

Nice :)! I got ÝεDÈi®*}}Ÿ without checking, ā®sm is crazy smart haha.

— Magic Octopus Urn

Answer 5

05AB1E, 15 14 bytes

ÝDÉ·<*Ý€û˜ÔsF¨

Try it online!

Explanation

Ý                # push range [0 ... n]
 D               # duplicate
  É·<            # (x % 2 == 1)*2-1 for each
     *           # multiply
      Ý          # range [0 ... a] for each
       €û        # palendromize each
         ˜       # flatten
          Ô      # connected uniqueified
           sF¨   # remove the last n elements

Answer 6

JavaScript (ES6), 56 bytes

f=(n,b=d=1,k=0)=>[k,...k-d*n?f(n,k-b?b:(d=-d)-b,k+d):[]]

Try it online!

Commented

f = (               // f = recursive function taking:
  n,                //   n = input
  b =               //   b = boundary value, initialized to 1
  d = 1,            //   d = current direction, initialized to 1
  k = 0             //   k = current sequence value, initialized to 0
) =>                //
  [                 // update the sequence:
    k,              //   append the current value
    ...k - d * n ?  //   if |k| is not equal to |n|:
      f(            //     append the (spread) result of a recursive call:
        n,          //       use the original input
        k - b ?     //       if k has not reached the boundary value:
          b         //         leave b unchanged
        :           //       else:
          (d = -d)  //         reverse the direction
          - b,      //         and use a boundary of higher amplitude and opposite sign
        k + d       //       update k
      )             //     end of recursive call
    :               //   else:
      []            //     stop recursion and append nothing
  ]                 // end of sequence update

Answer 7

Haskell, 43 bytes

f n=scanl(-)0[(-1)^k|k<-[1..n],_<-[2..2*k]]

Try it online!

Computes the negated cumulative sums of the list [(-1)^k|k<-[1..n],_<-[2..2*k]], which is the first n rows of

[-1,
 +1, +1, +1,
 -1, -1, -1, -1, -1,
 +1, +1, +1, +1, +1, +1, +1…

Answer 8

Jelly, 11 9 bytes

²Ḷƽ-*0;Ä

Try it online!

How it works

²Ḷƽ-*0;Ä  Main link. Argument: n

²          Square; yield n².
 Ḷ         Unlength; yield [0, ..., n²-1].
  ƽ       Take the integer square root of each k in the range.
    -*     Compute (-1)**r for each integer square root r.
      0;   Prepend a zero.
        Ä  Accumulate; take the sums of all prefixes.

Answer 9

Haskell, 48 42 bytes

f n=0:[(-1)^i*x|i<-[0..n-1],x<-[1-i..i+1]]

Try it online!

Thanks to Οurous for -1 byte

Even though it's kind of obvious in hindsight, it took me a while to arrive at (-1)^i*x which is x when i is even and -x when i is odd. Previous iterations where:

(-1)^i*x
x-2*mod i 2*x
(-1)^mod i 2*x
[x,-x]!!mod i 2
(1-sum[2|odd i])*x

Comments

You can save a byte by using 1-i instead of -i+1 in the .. expression.

— Οurous

Answer 10

C# (.NET Core), 300 167 bytes

I've never done any of these before, but this one seemed fun. I see why people use those "golfing" languages as 167 seems way higher than some of the other answers. But, you gotta go with what you know.

static int[] f(int n){if (n==1) return new int[]{0,1};var a=f(n-1);return a.Concat(a.Skip(a.Length-(n-1)*2).Select(x=>-x)).Concat(new int[]{(n%2)!=0?n:-n}).ToArray();}

Try it online!

// Recursive Worker Function
static public int[] f( int n )
{
    // Start with the simple case
    if ( n == 1 ) return new int[]{0,1};

    // Recusively build off of that
    var a = f(n-1);

    // To be added at the end
    int[] b = { (n%2) !=0 ? n : -n };

    // Skip some based on length
    int s = a.Length - (n-1)*2;

    // With the rest, multiply by -1 and then append to the end
    // And append the part
    return a.Concat( a.Skip(s).Select( x => -x ) ).Concat( b ).ToArray();
}

Comments

You can make this a lot shorter if you only count the using statements and the function. This is allowed by default unless the challenge specifies it must be a full program (even if it did, you could shorten the containing class name).

— Οurous

Thank you! Thanks to your suggestion, I figured out the meaning of the "header" and "footer" sections of the TIO site. That cut my submission size in half!

— Darrin Cullop

Welcome to PPCG! (this looks like your first post.) Don't worry about the other languages, just try to be as good as possible in your language. / Tips: Remove unnecessary spaces. In C# you can remove all spaces surrounding symbols ([](){};.) (n-1)*2 is just 2*n-2 and with some rearrangement you can remove the parentheses there.

— user202729

Besides, != has precedence less than % so you can remove a pair of parens. And you can use >0 instead of `!=0, saves a byte.

— user202729

Also from me: welcome to PPCG! Tips for golfing in C# and Tips for golfing in all languages might be interesting to read through. :) As for some golfing tips: static int[] f(int n) can become f=n=> by using a (recursive) lambda, and (n-1)*2 can become ~-n*2 to save on the parenthesis. I got it down to 155 (137 + 18) bytes: Try it online. The 18 bytes are for using System.Linq;, because required imports are mandatory for the byte-count. Enjoy your stay!

— Kevin Cruijssen

Answer 11

J, 25 bytes

-5 bytes thanks to FrownyFrog!

>:@*:$i.;@(<@i:@*_1&^)@,]

Try it online!

J, 30 bytes

>:@*:{.;@([:(i:@*_1&^)&.>i.,])

Explanation:

i.,] creates list 0..n

&.> for each number in the list execute the verb in (...) and box the result (I need boxing because the results have different length)

[:( _1&^) find -1 to the ith power (-1 or 1)

i:@* make a list -n..n or n..-n, depending on the sign of the above

;@ unbox

>:@*: find n^2 + 1

}. and take so many numbers from the list

Try it online!

Comments

would you consider writing the same code as a zero based n version? e.g *:{.;@([:(i:@*_1&^)&.>i.) .. the specification allows that

— jayprich

"n = 1 would then give 0 1 0 -1 -2"

— FrownyFrog

@FrownyFrog - Hmm, I didn't check it. I reverted to my first solution. Thank you for the observation!

— Galen Ivanov

25 Use $ for the cut-off, no need for &.> because * is rank-0.

— FrownyFrog

Answer 12

Python 2, 65 56 bytes

r=k=0;exec'print r;r+=1-k**.5//1%2*2;k+=1;'*-~input()**2

The output format is a bit ugly. :/

Try it online!

Answer 13

Java 8, 85 83 79 bytes

n->{for(int p=0,i=0;i<=n*n;p+=1-(int)Math.sqrt(i++)%2*2)System.out.println(p);}

-6 bytes thanks to @OlivierGrégoire.

Try it online.

Explanation:

n->{                            // Method with integer parameter and no return-type
  for(int p=0,                  //  Set both `p` to 0
      i=0;i<=n*n;               //  Loop `i` in the range [0, `n*n`]
      p+=                       //    After every iteration, increase `p` by:
         1-                     //     1, minus:
           (int)Math.sqrt(i++)  //     The square-root of `i`, truncated to its integer
           %2*2)                //     Modulo 2, and multiplied by 2
     System.out.println(p);}    //   Print integer `p` with a trailing new-line

Comments

Nice approach. I was working on such approach right now, to improve my answer, but you beat me to it (despite your meeting), well done! ;-)

— Olivier Grégoire

83 bytes (I just removed j).

— Olivier Grégoire

79 bytes: I made i go up instead of down to remove a redundant n*n.

— Olivier Grégoire

Hi. Writing this to inform me you that I basically ripped off your answer. (port to JavaScript). Hope it's ok

— Muhammad Salman

@MuhammadSalman Sure, np. I port answers from others pretty often as well. :) As long as the original answer is mentioned, like you did, it's all fine by me.

— Kevin Cruijssen

Answer 14

R, 48 46 42 bytes

for(i in 1:scan())F=c(F,-(-1)^i*(2-i):i);F

Try it online!

A port of the Ruby answer by Kirill L. - and saved 6 bytes thanks to the same Kirill L.! Now shorter than Giuseppe's solution ;)

A port of this Octave answer by Luis Mendo using approx is less golfy. n=n^2+1 can be replaced by ,,n^2+1; or by 0:n^2+1(positional argument xout) for the same byte count :

R, 56 bytes

f=function(n)approx((0:n)^2+1,-(-1)^(0:n)*0:n,n=n^2+1)$y

Try it online!

Comments

I think approx will work here in a similar manner to Luis Mendo's Octave solution as well.

— Giuseppe

@Giuseppe Thanks! It does work though it's longer. I've learned diffinv and approx from this question...

— JayCe

Although I also don't know a golfier way to do -1 power (in R ~ doesn't work as a complement operator :(), you can still save another 2 bytes by switching to a full program.

— Kirill L.

...and since it is a full program we can also use and spoil a predefined built-in: 42 bytes - finally, shorter than Giuseppe's!

— Kirill L.

Answer 15

APL (Dyalog Unicode), 17 bytes

+\0,¯1*⍳(/⍨)1+2×⍳

Try it online!

Golfed 2 bytes thanks to @FrownyFrog by converting to a train. See the older answer and its explanation below.

---

APL (Dyalog Unicode), 19 bytes

+\0,∊⊢∘-\⍴∘1¨1+2×⍳⎕

Try it online!

(Uses ⎕IO←0)

My first approach was to construct multiple ranges and concatenate them together, this easily went over 30 bytes. Then I started analysing the sequence

      +\⍣¯1⊢0  1  0 ¯1 ¯2 ¯1  0  1  2  3  2  1  0 ¯1 ¯2 ¯3 ¯4
0 1 ¯1 ¯1 ¯1 1 1 1 1 1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1

^{+\⍣¯1 denotes the inverse cumulative sum}

There is a repeating pattern of 1s and ¯1s, where the length of each consecutive sequence of 1s or ¯1s is 1+2×n. And each subsequence alternates between 1 and ¯1. What I can do now is to create the 1s and ¯1s list, and then scan by +

      ⍳4 ⍝ creates range 0..4
0 1 2 3
      2×⍳4
0 2 4 6
      1+2×⍳4
1 3 5 7
      ⍴∘1¨1+2×⍳4 ⍝ for-each create that many 1s
┌─┬─────┬─────────┬─────────────┐
│1│1 1 1│1 1 1 1 1│1 1 1 1 1 1 1│
└─┴─────┴─────────┴─────────────┘
      ⊢∘-\⍴∘1¨1+2×⍳4 ⍝ alternate signs
┌─┬────────┬─────────┬────────────────────┐
│1│¯1 ¯1 ¯1│1 1 1 1 1│¯1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1│
└─┴────────┴─────────┴────────────────────┘
      ∊⊢∘-\⍴∘1¨1+2×⍳4 ⍝ flatten
1 ¯1 ¯1 ¯1 1 1 1 1 1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1
      0,∊⊢∘-\⍴∘1¨1+2×⍳4
0 1 ¯1 ¯1 ¯1 1 1 1 1 1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1 ¯1
      +\0,∊⊢∘-\⍴∘1¨1+2×⍳4 ⍝ cumulative sum
0 1 0 ¯1 ¯2 ¯1 0 1 2 3 2 1 0 ¯1 ¯2 ¯3 ¯4

Comments

Checking other answers now, I see many also use the +\ method, but generate the sequence of 1s and ¯1s with ¯1*⌊.5*⍨×⍨⍳ which happens to be at least 3 bytes shorter.

— Kritixi Lithos

+\0,¯1*⍳(/⍨)1+2×⍳ is 17

— FrownyFrog

I knew my solution felt long

— Zacharý

Answer 16

Haskell, 47 bytes

(#0)
m#n|m>n=[-n..n]++map(0-)(m#(n+1))|1>0=[-m]

Try it online!

Answer 17

Java (JDK 10), 98 bytes

n->{var s="0";for(int i=0,r=0,d=1;i++<n;s+=" "+r,d=-d)for(r+=d;r!=i&r!=-i;r+=d)s+=" "+r;return s;}

Try it online!

Comments

Ah, while I was in my meeting you sneaked in an answer before me.. ;) Will leave mine as well through, because we use a completely different approach. +1 either way.

— Kevin Cruijssen

Answer 18

MATL, 17 15 bytes

-2 bytes thanks to Luis Mendo!

0i:oEqG:EqY"Ysh

Try it online!

Explanation for n=3:

0		% push 0
 i:		% read input as integer, push range
		% stack: [0, [1 2 3]]
   o		% modulo 2, stack: [0, [1 0 1]]
    Eq		% double and decrement, stack: [0, [1 -1 1]]
      G:	% push input and range again
		% stack: [0, [1 -1 1], [1 2 3]]
        Eq	% double and decrement,
		% stack: [0, [1 -1 1], [1 3 5]]
	  Y"	% run-length decoding
		% stack: [0, [1 -1 -1 -1 1 1 1 1 1]]
	    Ys	% cumulative sum
		% stack: [0, [1  0 -1 -2 -1  0  1  2  3]]
	      h	% horizontally concatenate
		% end of program, automatically print the stack

Answer 19

Octave, 44 42 41 bytes

2 bytes removed thanks to @StewieGriffin, and 1 byte further removed thanks to @Giuseppe!

@(n)interp1((t=0:n).^2,-t.*(-1).^t,0:n^2)

Try it online!

Answer 20

Ruby, 52 47 bytes

f=->n{n<1?[0]:f[n-1]+(2-n..n).map{|x|-~0**n*x}}

Try it online!

Below is the original 52-byte version with an explanation:

f=->n{n<1?[0]:f[n-1]+[(r=*2-n..n).map(&:-@),r][n%2]}

Try it online!

Walkthrough

f=->n{           #Recursive approach
 n<1?[0]         #Init with 0 if n=0
 :f[n-1]         #else make a recursive call
 +               #and append an array of numbers
 [(r=*2-n..n)    #Init r as splatted range from 2-n to n
 .map(&:-@)      #"-@" is unary minus, so this a fancy way to do map{|x|-x} for -1 byte
                 #For even n use this negated r, e.g. for n=4: [2, 1, 0, -1, -2, -3, -4]
 ,r]             #For odd n use r directly, e.g. for n=3: [-1, 0, 1, 2, 3]
 [n%2]           #Odd/even selector
}

Comments

I don't know Ruby - can you explain what this does especially the map(&:-@) portion?

— JayCe

@JayCe Added an explanation. Basically, this is just negation, what in R would be simply -r.

— Kirill L.

Thanks for the explanation - it helped me port this to R.

— JayCe

Answer 21

Prolog (SWI), 113 bytes

0-I-[0|I].
N-[H|T]-R:-N is -H*(-1)^N,A is N-1,A-[H|T]-R;I is H-(-1)^N,N-[I|[H|T]]-R.
N-O:-X is -N*(-1)^N,N-[X]-O.

Try it online!

Answer 22

Python 3, 83 bytes

def c(n):print([(-1)**j*(abs(j-i)-j)for j in range(n+1)for i in range(2*j)][:-n+1])

Answer 23

Husk, 18 17 bytes

:ṁoṡ₁ŀ⁰_₁⁰
*^⁰_1⁰

Try it online!

Answer 24

Charcoal, 19 bytes

Ｆ⊕ＮＩ×∨﹪ι²±¹…·∧ι⁻²ιι

Try it online! Link is to verbose version of code. Explanation:

  Ｎ                 Input as a number
 ⊕                  Increment
Ｆ                   Loop over implicit range
                ²   Literal 2
                 ι  Current index
               ⁻    Subtract
              ι     Current index
             ∧      Logical And
                  ι Current index
           …·       Inclusive range
       ι            Current index
        ²           Literal 2
      ﹪             Modulo
          ¹         Literal 1
         ±          Negate
     ∨              Logical Or
    ×               Multiply
   Ｉ                Cast to string and implicitly print

Alternative explanation:

Ｆ⊕Ｎ

Loop over the integers from 0 to the input inclusive.

Ｉ

Cast the results to string before printing.

×∨﹪ι²±¹

Negate alternate sets of results.

…·∧ι⁻²ιι

Form a list from the previous index to the current index, excluding the previous index.

Answer 25

Jelly,  11  12 bytes

^{Bah, I thought I had 11 wih _2+ỊrN)N;¥/}

_2+ỊrN×-*$)Ẏ

Try it online!

How?

_2+ỊrN×-*$)Ẏ - Main Link: n           e.g. 4
          )  - for x in [1...n]:           1       2          3               4
_2           -   subtract 2 from x        -1       0          1               2
   Ị         -   is x insignificant?       1       0          0               0
  +          -   add                       0       0          1               2
     N       -   negate x                 -1      -2         -3              -4
    r        -   inclusive range          [0,-1]  [0,-1,-2]  [1,0,-1,-2,-3]  [2,1,0,-1,-2,-3,-4]
         $   -   last two links as a monad:
       -     -     minus one              -1      -1         -1              -1
        *    -     raised to the power x  -1       1         -1               1
      ×      -   multiply                 [0,1]   [0,-1,-2]  [-1,0,1,2,3]    [2,1,0,-1,-2,-3,-4]
           Ẏ - tighten                    [0,1,0,-1,-2,-1,0,1,2,3,2,1,0,-1,-2,-3,-4]

Answer 26

Stax, 10 bytes

╗)SΘ█☼₧ΘP(

Run and debug it

Answer 27

Scala, 119 Bytes

def a(n: Int)={lazy val s:Stream[Int]=0#::Stream.from(0).map{x=>s(x)+1 -2*(Math.sqrt(x).toInt%2)}
s.take(n*n+1).toList}

Ungolfed:

def a(n: Int)={
  lazy val s:Stream[Int]= 0#::Stream.from(0).map //Give the starting point and indexing scheme
  {
    x=>
    {
      val sign = 1-2*(Math.sqrt(x).toInt%2) //Determine whether we are adding or subtracting at the current index
      s(x)+sign
    }
  }
  s.take(n*n+1).toList //Take the desired values
}

This can probably be golfed much better, but I wanted a solution utilizing lazy Streams.

Answer 28

APL (Dyalog Unicode), 34 32 bytes

⎕CY'dfns'⋄-0,∘∊1↓¨2to/Sׯ1*S←⍳,⊢

Try it online!

Requires ⎕IO←0

-2 bytes thanks to @FrownyFrog

Comments

32 as a train

— FrownyFrog

Answer 29

Stacked, 44 bytes

[~>0\:2%\#,2*1-tr[...rep]flatmap,$sumonpref]

Try it online! It's been a while since I programmed in Stacked, but I think I still got it.

Alternatives

73 bytes: [0\|>:2%tmo*2 infixes[:...|>\rev...|>rev#,$#'sortby 1#behead]flatmap 0\,]

This goes with the "ranges from generated indices" approach used in my Attache answer. This proved to be pretty long, since Stacked has no builtin for reversed ranges nor collapsing. (That's what :...|>\rev...|>rev#,$#'sortby 1#behead does.)

53 bytes: [0\|>:2%tmo _\tpo#,tr[...rep]flatmap 0\,inits$summap]

...so I decided to go for an approach which instead finds the cumulative sum (inits$summap) over 1 and -1 repeated by the odd integers, as in the R answer.

46 bytes: [~>0\:2%\#,2*1-tr[...rep]flatmap,inits$summap]

...but I realized that the negative integers and the odd integers could be made in one go, by multiplying both generated arrays (the mod 2 values of the range and the range itself) by 2 then subtracting 1. This gives alternating 1s and -1s for the first range and the odd integers for the second!

44 bytes: [~>0\:2%\#,2*1-tr[...rep]flatmap,$sumonpref]

... and then I remembered I had a builtin for mapping prefixes. ^-^

Answer 30

Julia 0.6, 44 bytes

n->[(i%2*2-1)*[0:i;(n>i)*~-i:-1:1]for i=1:n]

Try it online!

Since OP mentions "the output format is flexible", this prints an array of sub arrays, eg. U(3) => [[0, 1], [0, -1, -2, -1], [0, 1, 2, 3]].

i%2*2-1 decides the sign of the current subarray - negative for even numbers, positive for odd.

[0:i;(n>i)*~-i:-1:1] is in two parts. 0:i is straightforward, the range of values from 0 to the current i. In the next part, ~-i:-1:1 is the descending range from i-1 to 1. But we want to append this only if we're not yet at the final value, so multiply the upper end of the range by (n>i) so that when n==i, the range will be 0:-1:1 which ends up empty (so the array stops at n correctly).

---

And here's a version that can support random access - the inner lambda here returns the i'th term of the sequence without having to have stored any of the terms before it. This one gives the output as a single neat array too.

49 47 bytes

n->map(i->((m=isqrt(i))%2*2-1)*(m-i+m^2),0:n^2)

Try it online!